img2img: Three Consistent Characters (+Face Detailer)


Updated:

Workflow Preview

Showcases (Image/Video)

This workflow allowes one to turn images with up to 3 uniquely defined characters into a generated one.

Load the image into the four separated loaders. Edit the masks in the three lower loaders. Write 3 prompts for the characters and one for the bacvkground then launch the generation. Apparently, 10+ generations needed to obtain an appealing result.

Nodes Detail

68 Nodes
Primitive Node Types
0
Custom Node Types
68
  • CR VAE Input Switch1
  • CR Apply LoRA Stack1
  • ConditioningSetMask3
  • ConditioningCombine3
  • Reroute7
  • ImageUpscaleWithModel1
  • PreviewImage6
  • SaveImage3
  • UltralyticsDetectorProvider1
  • SAMLoader1
  • VAELoader1
  • DetailerForEach3
  • CLIPTextEncode11
  • ImpactSEGSOrderedFilter3
  • GetImageSizeAndCount1
  • SEGSToImageList3
  • ImpactSimpleDetectorSEGS1
  • UpscaleModelLoader1
  • VAEEncode1
  • CR Aspect Ratio1
  • ControlNetLoader1
  • LoadImage5
  • LatentUpscaleBy1
  • ControlNetApplyAdvanced1
  • OpenposePreprocessor1
  • CheckpointLoaderSimple1
  • CR LoRA Stack2
  • UltimateSDUpscale1
  • KSamplerAdvanced1
  • VAEDecode1